A Fundamental Shift in How the World Uses the Internet
The way the world interacts with the internet is changing, and it is changing very fast. Artificial intelligence is the catalyst, and the companies that positioned themselves early are now reaping extraordinary rewards. The most recent earnings season made this vivid: Google reported a 63% year-over-year increase in cloud revenue, Microsoft posted growth in the 40% range, and Amazon delivered numbers north of 20%. These figures are not incremental — they reflect a structural reordering of how digital services are consumed and delivered. Companies that anticipated this shift years ago are now harvesting the fruits of that foresight, while those that hesitated are scrambling to catch up.
This dynamic plays out at every scale. Whether the question is about public markets or about building a private business, the same principle applies: those who planned for this transformation are doing well, and those who did not are increasingly visible by their absence from the leaderboards.
Why Startups Still Have a Window
A natural question follows: how can any entrepreneur compete with companies that casually generate hundreds of billions of dollars per year and command multi-trillion-dollar valuations? The answer is that direct competition is the wrong framing. Startups and large enterprises are operating on entirely different clocks.
Large corporations want to use AI, but most of them simply cannot move quickly. They have to address data governance, privacy, security, and internal change management before their teams can meaningfully adopt the technology. In many big organizations, those conversations have not even begun. Meanwhile, startup operators are already deploying autonomous agents and building their daily workflows around AI. The advantage is not infrastructure — that ground belongs to the hyperscalers and the so-called Magnificent Seven — but speed of learning and product iteration. If you are building products rather than data centers, you can be first to market with entirely new categories of capability.
The Question of ROI and Labor
The eye-watering capital expenditures from the hyperscalers can feel almost like monopoly money. Eventually, that capital must produce a return, and the most obvious lever has historically been the replacement of labor. But there is a deeper question: have we ever truly replaced the thinking aspect of work? The honest answer is no. Past technological revolutions automated execution, not judgment.
That helps explain the strange duality of the current moment. Large companies are conducting layoffs even as smaller companies are finally able to attract talent. We are, in fact, in a productivity boom — and historically, productivity booms have never produced fewer jobs over time. They have produced more. This may eventually change. There may come a day, perhaps decades from now, when so much is automated that work itself looks unrecognizable. But that day is not in the next few years. The near-term reality is overwhelmingly one of expansion, not contraction, for those positioned correctly.
The Rise of the "Superhuman" Worker
Even in industries being most disrupted by AI — media and marketing among them — well-run companies are growing rapidly and hiring aggressively. The reason is that AI makes their products better and more scalable. The kind of person they want to hire is not someone whose only skill is "using AI." It is someone with a real underlying craft — marketing, content creation, analysis, sales — who has layered AI on top of that craft to multiply their output by an order of magnitude. The shorthand for this is the "superhuman" employee: a person who can 10x what they produce because they know both the discipline and the tooling.
When new tools arrive that appear to threaten an entire job category — for example, financial-analyst tooling released by AI labs and search platforms — the headline reaction is always that the role is finished. The reality is more nuanced. Today's AI is not perfect, and it still requires the human touch to direct, validate, and refine its output. The opportunity is not to be replaced by AI; it is to be the human who knows how to use AI to elevate the work.
Advice for Those Entering the Workforce
The job market is genuinely difficult right now for new graduates, and the stress on their faces at career events tells the story. But the picture is not uniformly bleak — it is bifurcated. Layoffs are concentrated in project management and low-level execution roles. Meanwhile, hiring is robust for people who know how to wield AI: prompt engineers, data scientists, internal AI educators (one major payments company is reportedly hiring someone solely to teach its marketing team how to use AI), and engineers who have integrated AI into their workflow. Engineers who refuse to adopt AI are the ones whose demand is collapsing. A telling anecdote: an engineering candidate proposed being hired alongside two junior reports working under him, and the response was that this made no sense — AI should be doing that scaling.
The takeaway for anyone entering the workforce is to develop two skills rather than one. First, get genuinely good at a craft that has existed for a long time — content creation, financial markets, marketing, sales, whatever fits. Second, figure out how to apply AI to that craft. Combining these makes you highly employable. Lacking both makes the job hunt brutal.
The Next Frontier
The current trade is dominated by AI infrastructure: chips, memory, compute, data centers. That buildout shows no sign of stopping. Forward projections for next year's spending have been revised from roughly $950 billion to about $1.1 trillion, and as long as those numbers keep climbing, capital will continue flowing into every adjacent layer of the stack.
Energy is one of the largest unsolved bottlenecks. Most of the data centers being announced are still not online, and a significant portion — roughly half — of recently planned projects have been cancelled. The remaining build-out will require enormous power, which makes renewables, natural gas, and other generation technologies essential to the AI thesis. You cannot run the next era of computing without solving the electricity problem first.
Space is the next adjacent frontier. When a major private launch and satellite company eventually goes public, it will mark the start of a meaningful capital cycle around orbital infrastructure — itself heavily tied to data centers, energy, and connectivity.
Most importantly, the focus is going to shift, probably within the next 12 to 18 months, from hardware to products. Investors and operators alike will start asking what real, recurring, SaaS-like revenue streams emerge from all this infrastructure. Robo-taxis are likely to be one of the first large-scale answers — a category where AI, manufacturing, and hardware converge into a tangible consumer product. The competitive battle between autonomous-vehicle operators and incumbent ride-share platforms will be one of the defining commercial stories of the coming years. More broadly, the next great wave is the embodiment of AI in the physical world: robotics, autonomous systems, and machines that translate intelligence into action.
Closing Thought
We have spent the last cycle watching software-as-a-service capital migrate into infrastructure-as-a-service. That migration is not the end of the story; it is a stage in a longer evolution. The infrastructure being built today is the substrate for a generation of products that have not yet been launched. For investors, operators, and workers alike, the message is the same: the opportunity is enormous, but it belongs to those who see AI not as a threat to be feared or a gimmick to be dabbled in, but as a multiplier to be mastered.